High Throughput Phenotyping of Biomass Sorghum Using Ground and Aerial Imaging.

Conventional phenotyping methods are laborious and time consuming. With advances in aerial imaging platforms, quick and accurate estimation of plant growth and development parameters of biomass sorghum can aid breeders to target physiological traits in a timely manner. Here, our goals were (i) to evaluate the efficacy of multiple vegetation indices to estimate biomass sorghum growth, plant height, and yield and (ii) to compare aerial images with traditional methods for estimating physiological and agronomic characteristics of biomass sorghum. A three-location (two fields in Suffolk and one field in Warsaw) field study was conducted in Virginia in 2017. Treatments were 16 biomass sorghum hybrids planted in early June (single season) and early July (simulating double cropping with wheat). Sensor data obtained from RGB images were aerial intensity (AI), aerial hue (AH), aerial lightness (AL), aerial a*, aerial b*, aerial u*, aerial v*, aerial green area (AGA; % pixels with 60<hue<120), and aerial greener area (AGGA; % pixels with 80<hue<120), as well as aerial crop stress index (ACSI). Ground-based and aerial normalized difference vegetation index (NDVI) data were also collected. We reported NDVI data as in-season estimated yield (INSEY) shown as (i) INSEY_DAP [NDVI divided by days after planting (DAP)], and (ii) INSEY_GDD [NDVI divided by growing degree days (GDD)], and the inverse simple ratio (ISR; [1–NDVI]/[1+NDVI]) divided by DAP (INSEY_ISR) and compared them with NDVI data. Our preliminary results for single season sorghum at Warsaw indicated that ground-based NDVI could predict dry matter (DM) yield (R² = 0.74). There was a positive linear relationship between AGA and sorghum DM yield (R² = 0.71), similar to that of AGGA and sorghum DM yield (R² = 0.71) when plants were 30 cm tall at V6 growth stage. There was also a negative linear relationship between aerial u* and yield (R² = 0.72) and ACSI and yield (R² = 0.69). Among vegetative indices, AGA and ACSI were the best predictors of leaf area index (R² = 0.55). Detailed results of all locations and sesning dates will be presented.